iSolutions is working to generate an advanced visualization and machine learning platform targeted at understanding drilling rig parameter impacts on Rate of Penetration (ROP). The platform will allow for drilling rig parameters to be tuned to optimize ROP for wells on existing and nearby Pads in the intermediate section. Using drilling information’s from historical wells, the model will be trained to estimate ROP’s given various input settings for all formations intersected in the vertical portion of a drilling operation. The model will consider parameters such as weight on bit, mud rate, bit type, RPM etc.
An optimization function will be generated to provide drilling engineers with critical drilling parameter settings for each formation intersected in the drilling of the wells. These recommended settings can be used as a starting point to provide a basis for further optimization work. The optimization function will allow engineers to perform ‘what-if’ analysis of a particular set of parameters without having to alter the actual drilling plan of a well. Hereby the situation(challenge), complication and solution for the project above are illustrated:
Drilling interbedded formations is challenging …
• Drilling parameters are important and must be optimized
• Different formations have different geological and geo-mechanical characteristics
• Bit life is vital in estimating the ROP and needs to be captured in the drilling optimization effort
There is a gap in OT analytics tools that can be leveraged to facilitate drilling optimization:
What makes current drilling operations complicated?
- Geological formation effect on ROP
- Unknown drilling parameters (RPM, WOB, etc) for optimized drilling performance
- Limited benchmarking work
- Challenging gas or liquid injection ratios
- Formation pressure estimation
The solution platform will use 3 stages, first data visualization and dashboarding, second shallow analysis in which the linear regression analysis and cross plots are created and third, which is the most value add will be deep data analysis and machine learning.
Neural Network Application: Predict ROP
The Neural Networks for each formation are created and tuned against the data. All networks are fed to an ROP optimizer for best KPI analysis. The results show best input for any given formation for getting optimum ROP. The optimizer is bundled with cost per footage such that uneconomical values for RPM, WOB etc, will be filtered out.